|
Abstract:
|
Big supply chains of multinational corporations depend on efficient procurement processes to ensure the smooth functioning of operations. All products and services needed to keep the supply chain functioning must be procured in time and in a cost-optimal and business-compliant manner to maximize savings while maintaining strong supplier relationships. The core of procurement operations consists of two decision problems: 1) when to send an order to a supplier, and 2) the supplier selection problem. We develop a unified probabilistic model of procurement, comprising a model of the demand and an evolutionary model of the procurement process, using the Markov Decision Process framework, with a joint mechanism of the two operational decisions. We propose an online learning algorithm of optimal procurement (decision) policies based on Thompson sampling, and illustrate, in a simulation study, how online learning can deliver insights to procurement managers about how they can adapt their purchasing strategy according to changes in the dynamics of the environment where the supply chain is operating and generate more savings for the business.
|
